Motor Memory Expansion Shapes Micro-Scale Dynamics of Naturalistic Skill Learning

Real-world motor skills rely on the precise spatiotemporal coordination of action sequences, progressively refined through practice. The performance of naturalistic skills depends on generating such sequences by forming motor memories that, in part, develop across rest periods interspersed with practice. Yet a critical gap remains in our understanding of how learning unfolds at a micro-scale within individual practice trials, as opposed to across trials or sessions. The moment-by-moment behavioral dynamics and their neural correlates that give rise to naturalistic skill acquisition during ongoing performance are incompletely understood.

To address this gap, we conducted two experiments in which participants learnt naturalistic sequential keypress skills of different speeds. Within each practice trial, we reproducibly identified segments of high-initial-skill (HIS)—brief periods of elevated performance followed by skill drops—that persisted throughout training. The keypress content of HIS segments scaled with execution speed and systematically increased with practice, reflecting an expansion of keypress chunk content. Importantly, this expansion of chunk content continued even after overall performance plateaued, predicting skill as training progressed.

Neural analyses revealed elevated theta–gamma phase–amplitude coupling (θ/γ PAC) and beta (β)-band bursts during HIS segments. Hippocampal θ/γ PAC notably predicted HIS segment content, identifying a possible mechanism contributing to binding individual actions into dynamic sequential enlargement of motor memory representations during early learning. These findings, inconsistent with a fatigue-based explanation, unveil expansion of motor memory capacity as a crucial contributor to early performance gains during procedural learning, bridging early cognitive control and expert skill performance.